OpenAI safety researcher says 70% chance AI makes humans extinct in 3 years

The past week has been a busy one in the AI world. After building the most advanced AI models the world has seen, experts are worried that things may go out of our control. OpenAI safety researcher Marcus Williams has claimed that if AI regulation is not done, humanity may be facing extinction within the next three years.

On X, Williams pointed out that there was a need for regulation around AI models, or for AI companies to work together on slowing down frontier AI development if humanity wanted to avoid a catastrophe. “Unless there is AI regulation or a coordinated slowdown between labs, human extinction in the next few years seems very likely,” he wrote.

Replying to a comment under the same post, Marcus Williams warned that we were running out of time, predicting that there was a “70 per cent” chance of extinction “in the next three years.” Although he added, “I think regulation/slowdown is very possible.” OpenAI's Marcus Williams believes humanity could be extinct soon.

As companies like Anthropic and OpenAI have developed advanced models, it has also raised the risks of AI going rogue or against human interests. Previously, OpenAI revealed that about 700 AI agents went rogue and tried to hack US company Hugging Face.

At the same time, OpenAI has announced US government technology adviser Paul Christiano is joining the board of its non-profit foundation and its Safety and Security Committee. “There is a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control in the very near term,” he wrote on X.

Paul Christiano insisted that there was a need for the AI industry to work together to ensure that AI superintelligence aligns with the best interest for us humans. “If we build superintelligence without more robust alignment I expect we will permanently lose control of it,” he added. “If that happens then most people could die.”

Superintelligence broadly refers to advanced AI that can match, if not surpass, human intelligence.

While Christiano discussed the need for joint efforts to slow down frontier AI development, he did not shy from acknowledging that even OpenAI needed to do more work to reduce risks. “I do not think that the AI industry in general, including OpenAI, is currently on track to reduce this risk to an acceptable level,” he wrote.

Previously, OpenAI chief scientist Jakub Pachocki also urged “extreme caution” over the future of AI. He likened AI to an “alien mind” warning that no one is prepared for the consequences of a continued rapid rise in machine intelligence.

Concerns have also deepened at Anthropic. Its alignment science lead, Evan Hubinger, said there was more than a 10 per cent chance AI could “kill all humans” within the next decade. While former AI researcher Jacob Coxon resigned, saying neither Anthropic nor OpenAI was acting responsibly and that they were “gambling with our lives.

The warnings come after recent incidents involving AI agents. Apart from the Hugging Face incident mentioned earlier, thousands of OpenAI agents also hijacked a German website to use as a message board. Anthropic had also disclosed a January incident in which a version of its Claude model in training broke into third parties after its task could not be aborted.- EndsPublished By: Armaan AgarwalPublished On: Sep 11, 2026 09:58 ISTAlso Read | Bioweapons, surveillance, espionage: Anthropic reveals how countries & groups are misusing ClaudeAlso Read | No longer merely aiding humans, but on course to become them: Ex-Anthropic researcherAlso Read | 10% chance AI kills humans next decade: Anthropic safety lead after colleague resigns